USING RETROSPECTIVE BEHAVIORAL-DATA TO DETERMINE HIV RISK FACTORS AMONG STREET-RECRUITED DRUG INJECTORS

This paper investigates whether the time period during which retrospec tive behavioral data are collected affects the results of studies of H IV risk factors. In particular, we address the concern that questionna ires probing behaviors in the recent past may not isolate the risk beh aviors involved at the time of actual HIV infection. During 1987 and 1 988, 278 street-recruited injecting drug users (IDUs) in New York City were interviewed about their sexual behaviors, medical history, drug- injection behaviors in the prior ten years, and specific needle-use be haviors during 1985-86 and thereafter. Among the 278 subjects, 52% wer e HIV-seropositive. In univariate analysis, serostatus was significant ly related to: a) total drug-injection frequency during 1983-84 and 19 85-86; b) mean monthly total drug-injection frequency during 1977 to 1 987; c)cocaine injection frequencies during 1983-84 and 1985-86, and d uring the last thirty days, and d) injecting in shooting galleries and using previously used cookers during 1985-86. Significant predictors of seroprevalence in stepwise logistic regression were total drug and cocaine injection frequencies during; 1983-84, years of injection, res idence in the Bronx, not being Hispanic and history of any sexually tr ansmitted disease, The data confirm previous reports on behavioral ris k factors for HIV serostatus. Re-analysis using drug;injection frequen cy data for different time periods indicates that in the context of an ''older'' HIV epidemic like that in New York City, analysis of risk f actors may be less sensitive to time periods than had previously been feared, although it may be true that data for behaviors in ''the last thirty days'' cover a sufficiently atypical period as to produce misle ading results.